Self-driving cars are stuck at idle, while driver assistance accelerates
Judging from the news of recent times, you might think that the era of human-controlled automobiles and other vehicles is rapidly drawing to a close. Take a look, for example, at this sampling of headlines from just the past month (in no particular order save for how I came across them):
- After reportedly dropping more than $1.1B USD of its own money on autonomous vehicle development, Waymo (a company under the "umbrella" of Alphabet, which also "owns" Google) is reportedly poised to invest another $1B on Lyft (and this isn't even the first $1B investment in this technology sector; Ford spent that amount for a majority stake in Argo AI back in February, while GM dished out more than $1B to buy Cruise Automation last year and Intel dropped more than $15B to buy Mobileye).
- Meanwhile, a court date nears for Waymo and Uber, Lyft's primary competitor, to sort out Waymo's autonomous vehicle technology theft accusations.
- The U.S. House of Representatives has passed legislation intended to accelerate the adoption of self-driving technology, while the U.S. Department of Transportation has introduced new (voluntary) guidelines for self-driving cars.
- General Motors is gearing up to ramp an autonomous version of the Chevy Bolt into production, based on Cruise Automation technology
- Waymo and Intel are collaborating (and have apparently been quietly doing so for some time) on self-driving car technology development.
- Intel and BMW have also added Fiat Chrysler to their autonomous vehicle R&D partnership
- Qualcomm has just unveiled a cellular wireless data chipset intended for vehicle-based applications, including vehicle-to-vehicle communications and coordination
- Lyft and Drive.ai are gearing up to test autonomous taxis in the San Francisco Bay area
- And speaking of testing, Ford and Domino's Pizza are trying out "autonomous" (not really) delivery, while Ford and Virginia Tech are analyzing others' reactions to self-driving (not really) roadway neighbors.
- And here's a name we haven't heard about in this particular application space until recently: Samsung has obtained a self-driving car testing permit in California, and has also just launched a $300M autonomous driving investment fund, along with putting $90M into Austria-based TTTech.
I could go on; I have enough candidates to roughly double the length of the above list without breaking a sweat. But I think I've made my point.
Here's the thing though: As anyone in engineering already knows well from personal experience, while getting something to work sometimes as a single-unit prototype under controlled laboratory conditions is one thing, getting it to work consistently in high volumes under real world conditions is a far different and more difficult matter. And since we're talking about vehicles carrying human beings (and potentially colliding with other human beings), getting them to work consistently is rather important.
The trials being conducted in the above lists are trials, nothing more than that. The investments being made are long-term, or at least they should be; anyone expecting a rapid return is foolish. To quote another analyst in a recent write-up, "we need to stop pretending that the autonomous car is imminent" (the title as originally published, "The Autonomous Car Charade", was even more scathing).
Need more evidence to support my skepticism? How about the fact that my doubts were reportedly shared by many of Tesla's engineers, a number of whom left (admittedly, in some cases, probably headed for other autonomous vehicle startups) when Tesla decided to roll out Autopilot anyway, over their objections. And, as I wrote last time, we all know how well (not) that experiment has turned out. Or look at how dramatically Apple has reportedly scaled back its self-driving car aspirations...to an autonomous campus shuttle?
Don't get me wrong: no matter the near-term skepticism of some customers (and passengers), full vehicle autonomy is a matter of when, not if. But when won't be nearly as soon as some advocates are trumpeting. And it'll come in stages (again, remember, the links that follow are just from the last month's news!):
- Take, for example, the elderly. Short, repeated trips dominate their needs. They are no longer capable of driving themselves, due to degraded vision, reactions, etc. but they still yearn for independence.
- Or take industries where the profit motive is compelling the elimination of the human "pilot" from the equation. Taxis, for example; there's a reason why Lyft and Uber are both making big-time investments here. Or semi-trailer trucks, although as with taxis, labor unions are resisting the move big-time (in a malicious manner, some fear). Or ships, for both patrol and transport purposes. How about drones, both for defense and delivery? Gliders? Farm equipment? Forklifts, robotic carts and other autonomous devices in factories, warehouses and the like? Food and other product delivery services? The opportunities and possibilities are countless.
- Whether the concept will take off first in urban or rural environs is not entirely clear to me, although if I was a betting man I'd put my chips on the former. City routes tend to be shorter and better mapped out, but the higher density of other vehicles, pedestrians, bikes and motorcycles and the like also makes them more hazardous.
And even in the absence of full autonomy, vehicles have already made tremendous beneficial strides when it comes to ADAS (advanced driver assistance systems). SAE International Standard J3016 describes six levels of autonomy. Here they are, in summary form:
- 0: No Automation
- 1: Driver Assistance
- 2: Partial Automation
- 3: Conditional Automation
- 4: High Automation
- 5: Full Automation
Tesla's Autopilot system, for example, is generally considered to be a Level 2 system, although its manufacturer claims that latest-generation vehicles contain hardware capable of Level 5 full automation (software capability is a different story). The Level 1 ADAS that I'm talking about, and that's commonly found to varying degrees today even with mainstream and even some entry-level manufacturers and models, consists of features such as the following:
- Front and rear collision avoidance, which not only warns you of an impending impact but also if necessary overrides the brake and/or accelerator and steering to prevent a crash
- Lane departure warning, which alerts you when you're drifting into the lane next to you and, if there's something already there, also takes over to make evasive maneuvers
- Road sign monitoring, which alerts you to speed limit changes, upcoming construction and other warning notifications, and the like, and
- Driver monitoring, which warns you if you're dozing off, say, or you're spending too much time checking Facebook or Twitter on your smartphone instead of keeping your eyes on the road as you should.
Yeah, that last one's particularly creepy from a personal privacy standpoint, I admit. But just as with Apple's new FaceID biometrics-based login scheme, I suspect folks will quickly grow accustomed to it. And the first time your car shakes your steering wheel or seat to lift you out of your hot sun- or late night-induced slumber, you'll wonder how you ever drove without it.
Take a look, for example, at this promo video for Cadillac's upcoming CT6 with Super Cruise, scheduled to begin shipping this fall:
and this charming clip of a 97-year old experiencing a Tesla Model S for the first time:
And then sound off in the comments with your thoughts. Thanks as always in advance!
- Tesla backlash shows misunderstanding of reality
- Driver-aware ADAS – The next step towards autonomous vehicles
- ADAS takes greater control in 2015
- Tesla's Autopilot: setting expectations ... or not
- Automotive radar chipset bolsters ADAS appeal
- Devkit for vision-based ADAS
- Optimizing compilers for ADAS applications
—Brian Dipert is Editor-in-Chief of the Embedded Vision Alliance, and a Senior Analyst at BDTI and Editor-in-Chief of InsideDSP, the company's online newsletter.